Send each PR to the engineers who know the code
Assign code experts based on who has worked on and knows the changed files, request reviewers by directory, or spread reviews around to share knowledge.
6PRs in one queue, none with a reviewerA docs fix waits behind an auth change. The oldest has waited 2d 4h.
Approved automaticallysafe_changes.cm
docs/ · 3 files · +48 −12test/ · 2 files · +96Sent to code expertscode_experts.cm
src/ledger/ · 6 files · +214 −38src/api/ · 4 files · +132 −20Security reviewsensitive_files.cm
src/auth/ · 3 files · +126 −41src/billing/ · 2 files · +58 −14Trusted by the teams transforming software delivery
gitStream reads the .cm files in your repository and runs them whenever a pull request opens or gets a new commit. Start from the automation library, or write rules for how your team already works.
Ifa PR only changes docs or teststhenlabel it safe-changeapprove itsay why in a comment
automations:safe_changes:if:- {{ is.docs or is.tests }}run:- action: add-label@v1args:label: 'safe-change'- action: approve@v1- action: add-comment@v1
MRmaya-r pushed 2 commits
docs/onboarding.md+32−8docs/setup.md+14−4README.md+2The gitStream automation library covers the jobs that slow pull requests down, from finding reviewers to running CI. Use each automation as it is, or change it to fit how your team works.
Assign code experts based on who has worked on and knows the changed files, request reviewers by directory, or spread reviews around to share knowledge.
Label every PR with its estimated review time, flag PRs that skip tests or delete files, and link each one to its Jira ticket.
Approve docs, tests, formatting and one-line fixes automatically, or let a team approve changes to the directories it owns.
Require security review for sensitive files and a second approval for large PRs, and check PR titles and changelogs, across every repository at once.
Trigger GitHub Actions workflows based on what a pull request changes, its branch or its labels, so a docs fix doesn't wait on your full test suite.
Merge Dependabot and Snyk updates once checks pass, approve code that clears your SonarCloud quality gate, and post to Slack or Microsoft Teams when a rule needs someone.
Reviewers open pull requests to a clear summary and an AI review that's already done. And work from AI coding tools gets its own label, so you can see where it lands.
/gs descAdds a retry budget to ledger sync, so a sync that keeps failing stops after five attempts and logs why.
Customers
Routing reviews across a shared monorepo
Pull requests are really where those different workstreams meet, and things can easily get tangled or delayed. Using the insights and the workflow automations we've set up helps us see where reviews might be stuck, route things smarter, and just get code merged more smoothly between the different groups.
Code experts on every pull request
The first gitStream rule we implemented – and the one that brought the most value – was auto-assigning PRs to people that were familiar with that area of the codebase already.
PR automation out of the box
We don't want to build custom tools unless we absolutely have to. LinearB automation gave us everything we needed right out of the box. Every time we run into a new PR problem, we ask: can LinearB solve this too? And so far, it's done exactly what we've needed.
GitHub, including GitHub Enterprise Server through a custom app, GitLab and Bitbucket Cloud. Most actions work on all three, with some exceptions: labels aren't available on Bitbucket, required approvals aren't available on GitLab, and comment commands and GitHub Actions triggers are GitHub only.
Minutes: about 2 to 5 on GitHub Cloud, 5 to 10 on GitHub Enterprise Server and 10 to 15 on GitLab or Bitbucket Cloud. You can install it on one repository, a few, or your whole organization.
Yes. When a pull request changes a .cm file, gitStream runs in dry run mode and comments with what each automation would have done, without doing it. You can also try rules against any GitHub pull request in the gitStream playground.
Yes. Rules can live in each repository's .cm folder, or in one organization-level cm repository that applies them everywhere.
Not on the Essentials plan. Managed Mode runs a set of ready-made automations, including AI review, AI descriptions, AI tool labels, review time labels, missing Jira labels and Dependabot approvals, from switches in LinearB. It works with GitHub. Custom rules live in .cm files in your repositories.
CODEOWNERS assigns reviewers by file path. gitStream rules can also look at what changed, who knows that code, how long the review will take and who wrote it, then label, approve, request changes, require more approvals or trigger a GitHub Actions workflow.
No. gitStream reads code and repository metadata only as needed to evaluate your rules and run AI features. It acts on the pull request, with labels, comments, reviewers, approvals and the merges your rules allow, and doesn't modify your code.
Yes. Rules can label pull requests from GitHub Copilot, Cursor, Claude Code and other AI tools, for example by the co-author on their commits. LinearB can then compare those pull requests with the rest on cycle time.
gitstream-cmbotcommented